NVIDIA has announced Rubin HPC, a new graphics processing unit in its Vera Rubin platform, aimed at scientific, engineering and analytics simulations. In a 16 November blog post published at the SC26 supercomputing conference, the company said the GPU would have “more than 50,000 [CUDA cores] on a single GPU,” a sparse-data acceleration capability intended to keep those cores supplied with data, and seventh-generation NVLink for GPU-to-GPU connectivity. Rubin HPC “is expected to be available in 2028.” Those figures and the date are NVIDIA’s.
The company framed the chip as a way to shorten the loop in which research agents propose candidate materials, molecules or designs and then rely on simulations to test them. “For agents tackling science and engineering, faster simulation means faster discovery,” said Ian Buck, vice president of hyperscale and high-performance computing at NVIDIA. “Rubin HPC accelerates the simulations agents use to test ideas and guide their next steps, helping scientists and engineers explore more possibilities and bring breakthroughs closer.”
NVIDIA said scientific workloads often share a pattern: an application reaches into a large dataset, gathers scattered values, calculates a result and repeats that at enormous scale. Rubin HPC, it said, is designed to scale data movement and compute together rather than adding compute alone, and to accelerate recurring simulation patterns including sparse matrix operations, stencil computations, sorting and reductions. Used with NVIDIA’s CUDA-X libraries, the company said, the GPU is intended to give agents a faster path through hypothesis, simulation and refinement. The blog listed climate models, engineering simulations, genomic analysis and financial risk as examples of the workloads it has in mind. Those are design claims for a product that is not yet shipping, and they are not accompanied by an independent benchmark or a customer installation.
The same-day SC26 announcements around it
On the same day, NVIDIA also announced a Vera CPU rack-scale platform for high-performance computing, combining 192 Vera CPUs with Quantum-X800 InfiniBand networking. The company said Vera CPU racks are expected to be available from Dell Technologies, HPE, Lenovo and Supermicro beginning in early 2027, earlier than the 2028 date given for Rubin HPC. It also introduced Nemotron for Science, a package of models, a chemistry dataset and post-training recipes for scientific agents. Those are separate announcements, not specifications of the Rubin HPC GPU.
The new GPU sits on a line NVIDIA has already been selling into the Vera Rubin generation, including Oracle’s launch of NVIDIA Vera Rubin on its cloud and earlier system work such as Dell’s Vera CPU racks. NVIDIA’s planned NVLink Fusion systems with Rebellions and SK Telecom are a neighbouring interconnect story, not this scientific GPU.
An announcement for 2028 is not a shipping GPU
The official record is a product announcement with an expected 2028 availability date, not a shipment, a price, a customer installation or an independent benchmark. “More than 50,000” CUDA cores is NVIDIA’s count. So are the claims about sparse-data acceleration, seventh-generation NVLink, and a balanced design that keeps those cores from waiting on data. Until independent measurements exist, those should be read as the vendor’s description of a chip that has not yet reached the market. The blog also does not give a price, a power rating, or a memory specification, so the public record is still a high-level architecture claim and a date.
The 2028 date also separates Rubin HPC from the same-day Vera CPU news. Vera racks, NVIDIA said, are expected from OEM partners in early 2027. Folding the two into a single “available next year” story would mix two different calendars. Nemotron for Science, likewise, is a software and dataset announcement, not a feature of the GPU itself.
What to watch next is not a launch event already in the diary, but any later specification, customer, or pricing disclosure that would move Rubin HPC from a conference announcement to a product with a confirmed shipment date.
Sources
- Dan Ernst, “NVIDIA Announces Rubin HPC — A New GPU to Accelerate Agents for Science, Engineering and Analytics,” NVIDIA blog, 16 November 2026
- NVIDIA, “NVIDIA Expands Vera CPU Platform to Accelerate Scientific Discovery”, 16 November 2026
- NVIDIA, “NVIDIA Introduces Nemotron for Science to Accelerate Discoveries With Agents”, 16 November 2026
Get TSN’s biggest tech stories, explained.
News and insights across AI, tech, crypto and cybersecurity. Every source linked. Unsubscribe anytime.
Get the newsletterRelated stories
- Oracle Launches NVIDIA Vera Rubin on Its Cloud With 150,000 Vera CPUs is the closest recent TSN account of the Vera Rubin generation moving into a customer cloud, and the right next stop for that platform file.
- NVIDIA, Rebellions and SKT Plan NVLink Fusion Systems
- Dell Packs 256 NVIDIA Vera CPUs Into One Rack, Adds Groq 3 Inference
Follow @tsnmediahub on X for breaking AI news as it happens.
Follow on X




